Traditional detectors only know the classes they were trained on. Open-vocabulary detection accepts a text prompt and locates whatever it describes, fusing a vision model with a language model. It is the practical way to ship detection for a long, changing list of objects.
The book · Volume 6 of 7
AI for Video: A Complete Guide to Intelligence in Video Products: From Models and Inference to Computer Vision, Speech, Search, and Generative Video
Open-vocabulary detection lets a product find new objects from a text prompt, which suits long or changing class lists but costs speed. This volume covers open-vocabulary detection in depth: Grounding DINO, Florence-2 and OWLv2, fixed-class detectors like RT-DETR, and deployment trade-offs. Written by Nikolay Sapunov, CEO at Fora Soft.
Is open-vocabulary detection right for your changing list of objects?
List the objects you need to find, and we will compare open-vocabulary detection with fixed-class detectors on accuracy and frames per second on your data. 250+ video products shipped since 2005.
